Transfer Learning of Decision Feedback Neural Network Equalizers for Faster-than-Nyquist Signals Transmitted over MCF

We investigate a transfer learning (TL) of decision feedback neural network equalizers (DFNNE) for Faster-than-Nyquist signals transmitted over multi-core fibers (MCF). In the simulation scenarios, the convergence rate of the TL-aid DFNNE with a source link of 17.5GHz is the fastest, achieving 72.2% faster compared to the traditional DFNNE.